Christophe wrote:chris wrote:Strange . . . Could you send or post the contents of shaders.log?
http://epicycle.free.fr/celestia-shaders.log
Ah . . . the combination of atmospheric scatter, cloud shadows, and the night texture map are using up all available interpolants. Interpolants are the values passed written in the vertex shader and passed on to the pixel shader. I can change a few things around so that this particular shader will run, but we'll run past the limit again if there's another light source. I need to add some code to use multiple passes--one per light source--in those cases.
chris wrote:It does? One of those images looks like a midday scene with a blue sky.And except at sunrise and sunset, the sky appears black.
You're right, the problem is not linked to the position of the Sun in the sky but to the altitude of the point you're looking at. The sky is always black at the zenith, even at midday.
That's another known problem . . . The atmospheric scattering model only accounts for single scattering, i.e. light that was only affected by a single scattering event on it's way from the light source to the viewer. Multiple scattering is going to be too slow to simulate in real-time. It's similar to the problem simulating multiple bounces of light between solid surfaces--in interactive rendering, it's almost always just direct lighting that is considered. My workaround will probably be similar: adding an ambient factor that's a catch-all for light from second and higher order scattering.
--Chris
It's still possible to end up with a shader that's too complex, however. I've got some other things I'll be changing to reduce the number of interpolators.
in awe at the stunning results of this new code or trying to understand this same code to see if I can get a clue as to how it works.
